v7.0 Breaking Changes

This is a list of planned breaking changes in v7.0.0.

• r, s values will be returned QUANTITY encoded for eth_getTransactionByHash
• eth_sign should return chainId-relative signature with high-level v values
• fix bug where ganache treats all transactions as having a chainId of 1, but RPC eth_chainId returns 1337 (we need to just make chainId configurable). We need investigate aligning net_version and chain_id as metamask uses net_version instead of chain_id for signing transactions.
• move reporting of tx hash on error to error field to prevent poorly-written clients which assume that the existence of the “result” field implies no errors from breaking.
• relative to above: don’t return both result and an err, ever. Maybe. Need to discuss further.
• flip the VM errors on RPC response flag
• old db format will no longer be valid. Investigate how to update old dbs to the new one without forcing the user into the new release.
• Update web3_clientVersion to include Ganache in its return value.
• improve precision of internal timeAdjustments (this is more of a bug fix that could break test that use specific block times via the time option, evm_increaseTime, or evm_setTime)

[davidmurdoch]

@frangio It was merged into develop. We’re working on buttoning some things up and porting the last feature: forking.

[MicahZoltu]

I am a big fan of the Robustness Principle. However, in JSON-RPC the presence/non-presence of those two variables are the intended mechanism for discriminating between success and failure, and having both is considered an error (per the specification). Because of that, I don’t think the robustness principle applies here as the correct way to discriminate is to look for either a success or an error and expect never both.